Task-Specific Summarization

Summarize source text for a fixed downstream question so retained details match the later prompt — trading reusability for relevance.

General summaries keep what seems “central” and drop asides — often exactly the asides a later app needs (a vacation post’s book mention for a recommendation agent). Task-specific summarization states the downstream use in the summarizer prompt (few-shot notes of what to jot vs N/A) so lossy compression keeps app-relevant signal.

Tradeoff: powerful when the feedforward question is stable across instances; if the question changes, re-summarize from scratch. General summaries are reusable across apps and models. Prefer task-specific inside dynamic context gathering when you know the ask; use hierarchical summarization first when the raw corpus will not fit even one summarizer call.

---

Sources

• Prompt Engineering for LLMs (Berryman), ch. 5